Original and REGICOR Framingham Functions in a Nondiabetic Population of a Spanish Health Care Center: A Validation Study

Francisco Buitrago; Juan Ignacio Calvo-Hueros; Lourdes Cañón-Barroso; Gerónimo Pozuelos-Estrada; Luis Molina-Martínez; Manuel Espigares-Arroyo; Juan Antonio Galán-González; Francisco J. Lillo-Bravo

doi:10.1370/afm.1287

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1
Risk category distribution of the population according to the original and REGICOR Framingham risk functions.

View popup

Table 1

Baseline Characteristics of the Cohort of Patients Studied

Characteristic	Overall Population (n=447)	Men (n=201)	Women (n=246)	P Value
Age, mean (SD), y	52.4 (9.6)	50.9 (9.8)	53.6 (9.2)	<.01
SBP, mean (SD), mm Hg	132.0 (17.8)	130.7 (16.6)	133.1 (18.7)	.147
DBP, mean (SD), mm Hg	82.1 (10.2)	82.5 (10.8)	81.7 (9.7)	.466
Arterial hypertension, n (%)	122 (27.3)	51 (25.4)	71 (28.9)	.410
Total cholesterol, mean (SD), mg/dL	238.9 (34.7)	239.3 (34.1)	238.7 (35.2)	.861
HDL cholesterol, mean (SD), mg/dL	53.6 (15.7)	47.3 (13.2)	58.7 (15.7)	<.001
LDL cholesterol, mean (SD), mg/dL	160.4 (33.2)	161.9 (31.9)	159.1 (34.2)	.870
Triglycerides, median (SD), mg/dL^a	107.0 (79.0–149.5)	131.0 (96.3–177.8)	96.0 (71.5–125.5)	<.001
BMI, mean (SD), kg/m²	27.6 (4.1)	27.5 (3.9)	27.7 (4.3)	.704
Smokers, n (%)	123 (27.5)	90 (44.8)	33 (13.4)	<.001
Ex-smokers <1 y, n (%)	20 (4.5)	16 (8.0)	4 (1.6)	<.01
GFR estimated by MDRD equation, mean (SD), mL/min/1.73 m²	76.8 (14.5)	80.9 (15.9)	73.4 (12.3)	<.001
GFR estimated by Cockcroft-Gault formula, mean (SD), mL/min/1.73m²	86.4 (21.2)	93.4 (21.2)	80.9 (19.5)	<.001
Patients with GFR <60 mL/min/1.73 m² by MDRD equation, n (%)	30 (6.7)	8 (4.0)	22 (8.9)	<.05
Patients with GFR <60 mL/min/1.73 m² by Cockcroft-Gault formula, n (%)	28 (6.3)	9 (4.5)	19 (7.7)	.158
Risk estimated by original Framingham function, mean (SD), %	11.6	16.2	7.8	<.01
Risk estimated by REGICOR Framingham function, mean (SD), %	4.3	5.7	3.3	.246
Coronary events, n (%)^b	30 (6.7)	22 (10.9)	8 (3.3)	<.01

BMI = body mass index; DBP = diastolic blood pressure; GFR = glomerular filtration rate; HDL = high-density lipoprotein; LDL=low-density lipoprotein; MDRD=Modification of Diet in Renal Disease; SBP = systolic blood pressure.
↵a Median of quartile 1–quartile 3, nonnormal distribution.
↵b Events during 10-year follow-up.

View popup

Table 2

Discrimination, Calibration, and Validity Statistics for Predicted 10-year Risk of Cardiovascular Disease by REGICOR and Original Framingham Risk Functions

	Original Framingham Function			REGICOR Framingham Function

	Men	Women	Total	Men	Women	Total
AUROC (95% CI)	0.63 (0.52–0.75)	0.65 (0.46–0.85)	0.71 (0.61–0.80)	0.63 (0.52–0.75)	0.65 (0.46–0.85)	0.69 (0.60–0.79)
Brier score^a	0.100366	0.033641	0.06364	0.097844	0.030775	0.06093
Predicted %/observed risk % (ratio)	16.9/10.9 (1.48)	7.8/3.3 (2.36)	11.6/6.7 (1.73)	5.7/10.9 (0.52)	3.3/3.3 (1.0)	4.3/6.7 (0.64)
Sensitivity, % (95% CI)	40.9 (20.4–61.4)	12.5 (0.0–35.4)	33.3 (16.5–50.1)	18.2 (2.1–34.3)	14.3 (0.0–38.5)	16.7 (2.4–30.0)
Specificity, % (95% CI)	72.1 (65.5–78.6)	97.5 (95.5–99.5)	86.6 (88.3–89.9)	87.2 (82.3–92.1)	98.7 (97.3–100)	93.8 (91.4–96.1)
Positive predictive value, % (95% CI)	15.3 (6.1–24.4)	14.3 (0.0–40.2)	15.2 (6.5–23.8)	14.8 (1.4–28.2)	25.0 (0.0–67.4)	16.1 (3.2–29.1)
Negative predictive value, % (95% CI)	90.8 (86.1–95.6)	97.1 (94.9–99.2)	94.8 (92.5–97.0)	89.7 (85.1–94.2)	97.1 (95.0–99.2)	94.0 (91.7–96.3)
Positive likelihood ratio	1.5	4.9	2.5	1.4	11.2	2.7
Negative likelihood ratio	0.8	0.9	0.8	0.9	0.9	0.9
Utility	1.8	5.5	3.2	1.4	13.4	3.0

AUROC = Area under the receiver operating characteristic curve; CI = confidence interval.
↵a A lower score indicates better accuracy of risk estimates.

View popup

Table 3

Characteristics of Patients Identified as of High Risk by the Original and REGICOR Framingham Functions

Characteristic	High Risk by Original Framingham (n=66)	High Risk by REGICOR Framingham (n=31)
Age, mean (SD), y	59.2 (8.6)	61.6 (7.1)
SBP, mean (SD), mm Hg	142.1 (15.7)	142.2 (11.5)
DBP, mean (SD), mm Hg	86.6 (10.4)	86.1 (7.6)
Total cholesterol, mean (SD), mg/dL	252.1 (32.7)	255.2 (30.5)
HDL cholesterol, mean (SD), mg/dL	42.3 (12.9)	40.2 (13.4)
LDL cholesterol, mean (SD), mg/dL	157.7 (32.0)	182.1 (33.7)
Triglycerides, median (SD), mg/dL^a	145.0 (105.0–206.8)	139.0 (106.0–205.0)
BMI, mean (SD), kg/m²	28.7 (4.4)	28.7 (3.8)
Smokers, n (%)	34 (51.5)	17 (54.8)
Ex-smokers <1 y, n (%)	10 (15.2)	5 (16.1)
Arterial hypertension, n (%)	25 (37.9)	11 (35.5)
GFR by MDRD equation, mean (SD), mL/min/1.73 m²	75.7 (10.4)	72.9 (8.8)
GFR by Cockcroft-Gault formula, mean (SD), mL/min/1.73 m²	85.7 (21.3)	79.9 (16.2)
Patients with GFR <60 mL/min/1.73m² by MDRD equation, n (%)	3 (4.5)	2 (6.5)
Patients with GFR <60 mL/min/1.73 m² by Cockcroft-Gault formula, n (%)	6 (9.1)	4 (12.9)
Coronary events, n (%)^b	10 (15.2)	5 (16.1)
Coronary risk estimated by original Framingham function, % (?)	28.5 (7.9)	34.2 (8.1)
Cardiovascular risk estimated by REGICOR Framingham function, % (?)	10.6 (3.4)	13.1 (3.6)
Men, n (%)	59 (89.4)	27 (87.1)

BMI = body mass index; DBP = diastolic blood pressure; GFR = glomerular filtration rate; HDL=high-density lipoprotein; LDL=low-density lipoprotein; MDRD=Modification of Diet in Renal Disease; SBP = systolic blood pressure.
↵a Median of quartile 1–quartile 3, nonnormal distribution.
↵b Events during 10-year follow-up.

View popup

Table 4

General Characteristics of the Patients With and Without Coronary Events

Characteristics	Patients Without Coronary Events (n=417)	Patients With Coronary Events (n=30)	P Value
Age, mean (SD), y	52.0 (9.5)	58.3 (8.8)	<.001
Male, no (%)	179 (42.9)	22 (73.3)	<.01
SBP, mean (SD), mm Hg	131.9 (18.0)	133.4 (16.2)	.657
DBP, mean (SD), mm Hg	82.2 (10.1)	80.9 (11.2)	.519
Total cholesterol, mean (SD), mg/dL	238.9 (34.4)	239.5 (38.9)	.617
HDL cholesterol, mean (SD), mg/dL	53.9 (15.8)	49.7 (13.0)	.160
LDL cholesterol, mean (SD), mg/dL	160.1 (32.7)	163.9 (39.3)	.558
Triglycerides, median (SD), mg/dL^a	107.0 (77.0–151.5)	107.0 (87.0–133.5)	.958
BMI, mean (SD), kg/m²	27.7 (4.1)	27.1 (4.1)	.573
Smokers, n (%)	111 (26.6)	12 (40.0)	.112
Arterial hypertension, n (%)	110 (26.4)	12 (40.0)	.105
Risk estimated by original Framingham function, mean (SD), %	11.1 (8.4)	18.4 (12.0)	<.01
Risk estimated by REGICOR Framingham function, mean (SD), %	4.2 (3.2)	6.9 (4.8)	<.01
GFR estimated by Cockcroft-Gault formula, mean (SD), mL/min/1.73m²)	86.7 (21.2)	80.9 (19.3)	.232
Patients with GFR <60 mL/min/1.73m² by Cockcroft-Gault formula, n (%)	26 (6.2)	2 (6.7)	.767
GFR estimated by MDRD equation, mean (SD), mL/min/1.73 m²	76.6 (14.3)	78.6 (16.8)	.483
Patients with GFR <60 mL/min/1.73 m² by MDRD equation, n (%)	26 (6.2)	4 (13.3)	.261

BMI = body mass index; DBP = diastolic blood pressure; GFR = glomerular filtration rate; HDL = high-density lipoprotein; LDL=low-density lipoprotein; MDRD=Modification of Diet in Renal Disease; SBP = systolic blood pressure.
↵a Median of quartile 1–quartile 3.

View popup

Table 5

Patients as Candidates for Drug Therapy According to the SCORE Guidelines Recommendations, With Risk Estimated by the Original Framingham and REGICOR Risk Functions

Recommended Therapy	Total (n=447)	Men (n=201)	Women (n=246)
Lipid-lowering therapy
Original Framingham, n (%)	64 (14.3)	57 (28.3)	7 (2.8)
REGICOR Framingham, n (%)	30 (6.7)	26 (12.9)	4 (1.6)
P value	<.001	<.001	.360
Antihypertensive therapy
Original Framingham, n (%)	56 (12.5)	41 (20.4)	15 (6.1)
REGICOR Framingham, n (%)	35 (7.8)	23 (11.4)	12 (4.9)
P value	<.05	<.05	.55

Additional Files

Figures
Tables

The Article in Brief

Original and REGICOR Framingham Functions in a Nondiabetic Population of a Spanish Health Care Center: A Validation Study

Francisco Buitrago , and colleagues

Background Risk prediction models are designed to estimate the probability of a patient developing a clinical condition based on known risk factors. This study evaluates the performance of 2 long-established risk scoring mechanisms for coronary disease: the original Framingham and REGICOR Framingham mechanisms.

What This Study Found Researchers find that one scoring mechanism overestimates risk, whereas the other underestimates it. This 10-year observational study of 447 adult nondiabetic patients in Spain finds that the Framingham risk function overestimates coronary risk by 73 percent, whereas the REGICOR Framingham function underpredicts the population�s coronary risk by 64 percent. Moreover, the original Framingham function selects a greater percentage of candidates for antihypertensive and lipid-lowering therapies than the REGICOR function. The proportion of patients included in the high coronary risk category also is doubled with the original Framingham equation.

Implications

That both models fail to accurately predict the population�s actual coronary risk in the 10-year follow-up period is not surprising to the authors. The original Framingham study was conducted before the widespread use of effective treatment for cardiovascular risk factors, so its equation currently overpredicts cardiovascular risk when applied to populations who have their risk factors actively managed.
The authors conclude the Framingham risk mechanisms could be improved by revising them to include additional cardiovascular risk factors and variables, such as family history of cardiovascular disease in a first-degree relative, social deprivation, body mass index, and current prescription of antihypertensive therapy.